2 minute read

Any time something’s broken on a team, I start with the same question: what’s the incentive structure that would cause this behavior to make sense?

Most problems aren’t mysterious once you look at it that way.

That’s why I liked Tobi Lütke’s AI memo at Shopify so much. He’s not just saying “AI is important.” He’s spelling out exactly how it should be used and then wiring the expectations directly into how the company operates. Here’s what he laid out:

  • Build your prototype with AI.
  • Your performance will reflect how well you’re using these tools.
  • You’ll get feedback from the people working next to you.
  • You can’t add headcount until you’ve shown that AI can’t handle it.

None of this is vague. These are concrete levers tied to real feedback loops and outcomes. It gives people room to learn, but it also makes learning the job.

And I think that’s the only way this works. There’s no shortcut here. No course or onboarding doc that will get your team fluent. People get better by trying. Repeatedly. It’s just reps.

I’ve seen firsthand how fast things change when one or two people inside a company start seriously experimenting. It shifts the energy. Suddenly people are sharing prompts, asking smarter questions, and finding real edge cases where these tools work better than expected.

Teams that make it easy to share what they’re learning — prompts, workflows, what failed — will compound faster. That kind of habit becomes a real edge.

But it only scales if leadership matches that bottom-up energy with real structural support and incentives that align with their vision for adoption. When those two things line up, it’s a huge differentiator.

The other part of this — and the thing I don’t think enough teams are paying attention to — is how fast the gap is going to widen. Some companies are slowly circling AI, waiting for a playbook. Others are already building with it every day. The difference won’t be obvious until suddenly it is.

I know that sounds like hyperbole. But I don’t think it is. The teams that figure this out aren’t going to just pull ahead — they’re going to make the rest look like they’re standing still.

If you’re leading a team, the question isn’t whether AI matters. It’s whether your team is actively building the muscle to use it well. That takes real reps, clear expectations, and incentives that reward progress. The teams that treat this seriously will build a compounding edge. And in this case, being a fast follower won’t be fast enough. This isn’t something you can afford to follow. You have to lead.

Updated: